Skip to content
master
Go to file
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
 
 
 
 
 
 
 
 
 
 

README.md

Partially-Aligned Data-to-Text Generation with Distant Supervision

This is the code for the EMNLP 2020 paper "Partially-Aligned Data-to-Text Generation with Distant Supervision". Traditional text generation task requires well-aligned data which is expensive to annotate. We relax the strict restrictions and propose this new task aiming at utilizing automatically made partially-aligned data. This method considerably expands the application domains where only automatically partially-aligned data is available.

drawing

Requirements

  • GCC >= 4.8
  • Python >= 3.7

Install

git clone https://github.com/fuzihaofzh/distant_supervision_nlg.git
cd distant_supervision_nlg
./scripts/setup.sh

Preprocess Data

./scripts/preprocess.sh wita50k

Train Baseline Model (Optional)

The model will be evaluated automatically during training.

# Train S2ST model
./scripts/train.sh wita50k base
# Check Score
tail -n 1 output/eval/wita50k__base/eval.100.txt

Train Our DSG Model

The model will be evaluated automatically during training.

# Step 1. SE Training
./scripts/train.sh wita50k endorsement,pretrain
# Step 2. S2SG Training
./scripts/train.sh wita50k endorsement,beam_endorse
# Check Score
tail -n 1 output/eval/wita50k__endorsement,beam_endorse/eval.100.txt

Cite

@inproceedings{fu2020partially,
  title={Partially-Aligned Data-to-Text Generation with Distant Supervision},
  author={Fu, Zihao and Shi, Bei and Lam, Wai and Bing, Lidong and Liu, Zhiyuan},
  booktitle={Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP)},
  year={2020}
}

About

Code for the paper "Partially-Aligned Data-to-Text Generation with Distant Supervision" in EMNLP 2020.

Resources

Releases

No releases published

Packages

No packages published
You can’t perform that action at this time.